Semanlink - Node Embeddings

Βρες μου:

Search Doc:

Προτιμήσεις...

Ετικέτες

ACL - AI@Google - Arxiv Doc - Attention in Graphs - Discuté avec M. Labeau - Entity Disambiguation - Jure Leskovec - KBPedia - KG-augmented Language Models - Neo4j - Olivier Ferret -

Γονείς:

Graph Embeddings

Node Embeddings

How do we do node embeddings? ([source](http://snap.stanford.edu/proj/embeddings-www/index.html#materials))

Intuition: Find embedding of nodes so that “similar” nodes in the graph have embeddings that are close together.

1. Define an encoder (i.e., a mapping from nodes to embeddings)
	- Shallow embedding (simplest encoding approach): encoder is just an embedding-lookup. Ex: [node2vec](/tag/node2vec), DeepWalk, LINE
2. Define a node similarity function, eg. nodes are similar if: 
	- they are connected?
	- they share neighbours?
	- have structural similar roles?
	- ... 
3. Optimize the parameters of the encoder so that similarity in the embedding space (e.g., dot product) approximates similarity in the original network

Defining similarity:

- Adjacency-based Similarity
- "Multihop" similarity (measure overlap between node neighborhoods)

these two methods are expensive.
-> **Random-walk Embeddings** (Estimate probability of visiting node v on a random walk starting from node u using some random walk strategy, optimize embeddings to encode random walk statistics). Expressivity (incorporates both local and higher-order neighbourhood information) and efficiency (do not need to consider all pairs when training)

Which random walk strategy?

- fixed-length random walks starting from each node: **DeepWalk** (Perozzi et al., 2013)
- "biased random walks" that can trade off between local and global views of the network: **Node2Vec** (Micro-view / marco-view of neighbourhood)

No method wins in all the cases

> We can now view the popular "node embedding" methods as
well-understood extensions of classic work on dimensionality reduction [src](doc:2020/08/graph_representation_learning_b)

Σχετικές ετικέτες:

Απόγονοι

1 Document (Όλα τα έγγραφα)

Watch Your Step: Learning Node Embeddings via Graph Attention

Tags:

2019-08-23 About

Ιδιότητες

sl:creationDate : 2018-05-10
sl:creationTime : 2018-05-10T14:33:26Z
rdf:type : sl:Tag
skos:prefLabel : Node Embeddings